Any datasets you have the rights to use are acceptable. Which dataset is it? If it's open for public use then you're probably fine. If it's private and was only shared with a previous group then you might not be. I'm happy to ask for you.
We tried really hard to find a definite answer, but could not find an official response because the content is outdated. That being said, the underlying issue is still relevant – building a way to ‘pull out messy, unstructured data from ads’ and/or other communications, as well as a way to sort, transform, store, and/or send that cleaned data to authorities or other relevant groups as a notification or in an accessible/legible/actionable way, would be a really amazing application/set of applications.